Examples

See how a model works—in your terminal.

Load an Ollama or GGUF model, step through its architecture, and inspect the inputs, outputs, tensor shapes, and math behind each operation.

From model to math

Start an Ollama model in your terminal:

byhand model --ollama llama3.2
byhand model view for Llama 3.2 showing tensor shapes, grouped attention blocks, and the selected K projection
01 Step through operations and tensor shapes
byhand operation explainer for the Llama 3.2 K projection showing its input, learned weights, output, and matrix multiplication
02 Inspect the operands and math behind a result

Select a screenshot to view it at full resolution.

HTML export

Example: llama3.2

Prefer a browser? Export the same visualization as an interactive HTML file:

byhand model --ollama llama3.2 --output html

Excel export

Example: llama3.2 workbook

Prefer a spreadsheet? Explore the formula-driven model pipeline in Excel Online:

byhand model --ollama llama3.2 --output excel --step 5